FR: Article Labeling & Source Quality Statistics

[koichiio]

What is that?

Allow the user to label articles (e.g., good/bad) and store them in the DB. Provide aggregated stats per source accessible via CLI and TUI, such as the following:

• rate: (bad articles / total articles per source)
• ratio: (good: bad: unlabeled)

Why is this useful?

RSS/Atom feeds are great for discovering and testing new sources. However, over time, it becomes difficult to track which sources consistently provide valuable content and which do not.

This feature would allow users to:

• Evaluate sources based on actual reading experience
• Identify low-value feeds
• Gradually prune subscriptions
• Reduce noise and time spent reviewing low-quality articles

So, it turns feed consumption into a measurable feedback loop.

---

I've been using Nom for a while and absolutely love this project. I wish I could write Go and could contribute.

[guyfedwards]

Thanks for the issue @koichincom. This is an interesting idea. Would some sort of statistics around tags work? E.g. you can create tags such as rate-1, rate-2 etc for whatever you want, but then to be able to see counts of those tags by feed source?

[koichiio]

Thanks for your reply! @guyfedwards
Assuming you're trying to generalize the idea by per-article tagging, it sounds great. Two things I wanna mention specifically for the quality stats are

• Tag should come with a percentage, and rate fields are also needed
• The numerator of the rates needs to be specified in the config, given that tags are flexible

source	loved_it	not_interested	low_quality	unlabaled	low_quality_rate
source_a	80 (20%)	80 (20%)	40 (10%)	200 (50%)	30% (120/400)
source_b	40 (5%)	40 (5%)	160 (20%)	560 (70%)	25% (200/800)

In this case, the numerator of low_quality_rate is low_quality and not_interested, and the denominator is the number of all articles from the source (const). You might find the ratio feature redundant tho